In The Plex

by Steve Levy

  In February 2006, Google appointed Larry Brilliant to head DotOrg. Brilliant was a charming man with a medical degree and a mind-blowing résumé that included a key role in eradicating smallpox and a close relationship with the Grateful Dead. DotOrg’s biggest successes were modest compared to its aspirations. It worked best when it tapped Google’s unique assets to take on a problem at scale. Its archetypical success story was Google Flu Trends, which data-mined the behavior of search users to quickly locate outbreaks of the disease. At a dinner during the TED Conference in 2009, one early Googler, Lori Park, approached the head of the Bill and Melinda Gates Foundation and asked for his opinion of Google’s efforts. Bill Gates said that DotOrg “is the most publicized foundation in the world, and it’s tiny. Expertise and analysis is this much of what’s needed.” He made a gesture with his thumb and index finger a half inch apart to indicate how insignificant that amount was. “You make an impact with money,” he continued, referring to DotOrg’s outlays, in tens of millions compared to his own foundation’s billions. “Your analysis won’t help sick people or save people’s lives! You do that with monnnn-ney.”

  In April 2009, Brilliant resigned his post and Google installed Megan Smith to head the division. She helped curtail DotOrg’s wildly ambitious agenda, promising to focus on the measurably useful projects such as Google Flu Trends and other ventures that leverage the company’s assets. Smith explained the change to Googlers at a TGIF. “Money really matters,” she said. “We don’t have the kind of money that Ford and MacArthur have. But they don’t have the engineering talent we have.”

  The corporate cut that most disturbed employees was Google’s first significant layoff, involving a hundred people in a recruiting operation in Phoenix. “I was always worried that day would come,” says Judy Gilbert. Though it made impeccable sense—hiring was stagnant, so who needed all those recruiters?—laying off Googlers simply wasn’t Googley. Page, Brin, and People Operations executives had to endure hostile questions at the TGIF interrogation after the layoffs, and they assured people that there would not be larger cuts to come.

  Google didn’t stop recruiting the best people it could find, especially engineers. In fact, the effort became more urgent because there were vacancies at Google created by valued employees who either joined tech firms that were newer and more nimble than Google or started their own companies. And every so often, an early Googler would simply retire on his or her stock-option fortune. The defections included high-ranking executives and—perhaps scarier to the company—some of its smartest young engineers. The press labeled the phenomenon Google’s “brain drain.” Sheryl Sandberg, who had built up the AdWords organization, left to become the chief operating officer at Facebook. Tim Armstrong left his post as head of national sales to become CEO of AOL. (“We spent all of Monday convincing him to stay,” said the grim Sergey Brin at that next week’s TGIF, expressing well wishes toward its valuable sales manager.) Gmail inventor Paul Buchheit joined with Bret Taylor (who had been product manager for Google Maps) to start a company called FriendFeed. Of the eighteen APMs—Google’s designated future leaders—who had circled the globe with Marissa Mayer in the summer of 2007, fewer than half were still with the company two years later. All of them left with nothing but respect and gratitude for Google—but felt that more exciting opportunities lay elsewhere.

  Bret Taylor, while specifying that he cherished his time at Google, later explained why he’d left. “When I started at the company, I knew everyone there,” he said. “There’s less of an entrepreneurial feel now. You have less input on the organization as a whole.” When he announced his departure, a procession of executives came to his desk asking him to reconsider. “I didn’t know Google had so many VPs,” he said. But he’d made his mind up.

  Google tried to respond. “As we shift from the crazy days of backing up the truck and hiring as many people as we can, we’re focusing more on career development,” said Judy Gilbert. At a TGIF in October 2009, Laszlo Bock tried to explain the new reality. “Googlers don’t care about microkitchens or how we pay,” he said. “It’s about how we think.” Bock elaborated on how his team was shifting focus. Whereas People Operations had previously concentrated on maintaining the overtaxed Google hiring machine, now it would concentrate on “keeping people happy.” And how would it do that? With data, of course. Just as Google supplied analytics to website owners and advertisers, People Ops would develop a set of metrics to generate data to “inform people decisions.” There would even be a “people analytics team.” Bock’s group would conduct experiments and simulations in areas such as interviewing, hiring, compensation, and performance. They would construct statistical analysis curves to determine factors influencing Google’s attrition rate.

  Judging from the questions from the Googlers in attendance at Charlie’s that day, the reaction was skeptical. One Googler complained that with the newly static workforce, the traditionally quick promotions were slowing. Sergey Brin remarked that since Google’s organization was so flat, promotions were always hard.

  That was something employees were well aware of. Left to figure out how to handle the complexities of a 20,000-person company—“Larry and Sergey definitely don’t want to talk about career ladders,” says Judy Gilbert—Google’s People Operations team had constructed a system with nine levels of employee status below the top executives (who were tens and elevens on that scale). Some of the distinctions were vague. Often, Google didn’t even share with employees what level they occupied on the ladder, an odd departure from its usual internal transparency. Bock would explain that the stealth was due to “cognitive heuristics.” These were the deep-seated mental processes that made people think that they should defer to someone with a higher title. “That might help you on the savannah and it might help you in big companies, but it doesn’t help us at Google,” says Bock. “Eric and Larry want anybody to be able to tell someone, ‘You’re wrong,’ and give ten reasons why.” Titles got in the way of that.

  In February 2008, Eric Schmidt sent word to Chad Hurley that it was time for YouTube to get more serious about the bottom line. As Hurley put it in an email that month, the unit was “redirecting our efforts from user growth to monetization.” The biggest personnel change was the arrival of Salar Kamangar in the San Bruno office, where he would spend “three and a half days a week” (as Kamangar would say in his usual clipped deadpan) on YouTube.

  Hurley had been bugging Salar to make the jump for a while—even in meetings before the acquisition, he felt there was kind of a glow of success around Kamangar, a result of his work in developing Google’s ad system. Hurley thought how awesome it could be if Salar could do the same thing at YouTube. Now that there was more urgency to make money, it was a perfect time for Kamangar to arrive. Best of all, even though internally he was almost as much a Google icon as Larry and Sergey, Kamangar also appreciated that YouTube worked best while at arm’s length from its parent.

  At the same time as Google was stepping up its efforts to make profits with YouTube, it argued against the common perception that the service was bleeding money. Some commentators were calling the purchase an outright blunder and comparing YouTube unfavorably with Hulu, a website that combined selected programming from its owners, several television networks and studios. (Hulu was closer to the Google Video concept than to YouTube.) Some analysts figured that Google was burdened by sky-high video-serving costs. One widely circulated report by Credit Suisse in April 2009 calculated that YouTube was spending more than $350 million a year to stream an estimated 75 billion video plays to users. Google would privately tell journalists that those guesses were based on what others had to pay to move such massive numbers of bits. With its superefficient cloud infrastructure and its private fiber-optic network, Google’s costs were less, much less. (Exactly how much less, the company wasn’t saying, but Ramp-Rate, another company conversant with infrastructure costs, made its own assessment of $83 million.) In addition, because of the combination of Moore
’s Law and Google’s infrastructure improvements, Kamangar would note that the costs of streaming always went down. “I think it’s halving every year,” he says. (Of course, Google could have silenced the critics by simply sharing the actual numbers; the congenitally secretive company chose not to do so.)

  It was trickier to manage the costs of licensing from studios and other content holders. “In order to get it now, in some cases, we have to do things that are unnatural, like offering guarantees we can’t expect to recoup,” Kamangar says. “But we’ve made some good trade-offs and brought the costs down, so that’s helping with profitability.” The key was both breaking the ice with content companies and making the deals. Kamangar was boggled when he began to unravel the complicated tapestry of rights, permissions, and claims that governed licensing agreements in Hollywood and in the music industry.

  Without music rights, millions of homegrown videos created by YouTube users violated copyright—an amateur director would use music from a personal collection as a sound track on a video, or sometimes the sound track would simply be music playing ambiently. (If you captured your child’s first steps on video and in the background a radio was playing a song, the entire clip infringed copyright.) Kamangar didn’t put a value judgment on the way the labels and studios worked but tried to crack their code, talking to executives, producers, agents, and managers. One day he happened to be in New York and was invited to meet with the CEO of Universal Music Group, Doug Morris. Kamangar was escorted by bodyguards to a private elevator and ushered to a fancy office high above the city. He couldn’t help thinking of the contrast with Google, where you stumbled in and went to the microkitchen for coffee. Kamangar didn’t dwell on the irony that it was the scruffy kids in shorts, munching energy bars and writing analytics programs, who were pushing aside the old power structure. While he put the pieces of YouTube together, though, he always kept in mind that he was documenting a traditional media system on the verge of collapse. He had to deal with the music world as it was but also plan for the way it would be after disruptions, which Google and YouTube were accelerating.

  Kamangar had some specific ideas for improvement of YouTube. He urged a simpler user interface and a smarter recommendation system to point users to other videos they might enjoy. He urged more flexibility with producers of professional video so YouTube would get more commercial content. He also emphasized how some of Google’s key attributes—notably speed—had a huge impact on the overall experience. If Google could reliably deliver videos with almost no latency, he reasoned, users might not balk so much at the “preroll” ads that come before the actual content, especially if the video was one of a series that users subscribed to and so were already eager to see what was coming.

  But maybe the biggest contribution that Kamangar made was putting an end to the “silver bullet” theory—that lurking in someone’s imagination was a multibillion-dollar idea that would enrich YouTube as dramatically as AdWords had transformed Google’s bottom line. Since Kamangar had cocreated AdWords, he was able to declare that no such equivalent existed and YouTube should develop a broader, multifaceted revenue strategy, making use of some of the concepts of Google’s ad model but hitting some corner shots as well.

  A lot of his ideas for monetization, though, had the spirit of AdWords. Just as with Google search keywords, sometimes it was appropriate to show relevant ads with videos, sometimes not. “If I’m watching a kite-surfing video, it’s very likely that I’d be interested in buying the board that the kite surfer is on or taking a lesson from that person,” he says. Taking advantage of this symbiosis would open the door for bigger advertisers selling sports equipment or bathing suits, as well as small, long-tail advertisers, such as kite-surfing teachers looking for students in their zip code.

  In addition, people uploading videos for free viewing might be willing to pay Google to promote them as sponsored links—a one-click connection would then appear alongside organic search results like an AdWords ad, in either the search results page or the results page from a YouTube search. YouTube also began experimenting with “interest-based” advertising, in which ads would be personalized to the subjects that users had previously accessed. (This would be something that privacy-conscious users could opt out of.) Finally, YouTube was exploring something that Google Video had tried without success: paid viewing for premium videos.

  It was essential, says Kamangar, for YouTube videos to find their place in an advertising-centric ecosystem. “If we don’t figure out how to advertise this correctly, we’re not going to bring users a lot of the content that they would want.” But if Google did figure this out, people would produce movies, shows, and clips that would never have otherwise existed, just as people made videos of their cats because YouTube provided a venue for airing such digital folk art. Kamangar was amazed that documentary filmmakers had to scramble for a measly million dollars to make a movie that could profoundly affect people. If YouTube could make it worth their while, there would be many more such documentaries! “The previous model was built on scarcity, where you see things in windows—the movie window, the DVD window, the cable window,” says Kamangar. “The Internet is completely different, where you expect to have everything available to you at all times. But you relate to other people based on the similarity of content that’s now so niche that you self-identify with that.”

  Personally, though, Kamangar was cautious in sharing his video likes—or uploading his own videos—with a wide community. “I’m kind of private and only want to share with people that I know,” he says. But that didn’t mean that he wasn’t an enthusiastic supporter of YouTube’s community. “There’s an aliveness to YouTube, a set of values that make it less of a platform machine and more of a living, breathing set of people.”

  Google also became more aggressive in connecting sponsors for popular videos. A paragon of YouTube’s business model was “Fred,” a video channel created by a Columbus, Nebraska, teenager named Lucas Cruikshank. The teen pretended to be a six-year-old kid named Fred Figglehorn in a series of two-minute videos. “Fred is the George Clooney of YouTube,” says Hunter Walk. “He was the first one with a million subscribers. He uploads videos, and we put ads against them. Sometimes he sells product placement ads. Fred makes a million dollars a year. He just signed a movie deal.” The Fred videos—generally manic rants in which Cruikshank portrays a hyperactive, possibly brain-damaged child who speaks like one of Ross Bagdasarian’s chipmunks—often sported commercial messages for sponsors such as Samsung, the Food Channel, and Bratz on an overlay at the bottom of the window. Since he started in 2008, at age fourteen, Fred’s YouTube videos have chalked up over half a billion viewings. Though Fred’s success was solely a product of YouTube, people in the company never met the phenom. “We sent him a cake once,” says Walk.

  YouTube helped Fred’s youthful creator not just by selling ads but by providing analytics, the same way it did for AdSense publishers. (This was a result of an initiative called the YouTube Insight project, developed by engineers in Google’s Zurich center.) Such data helped creators learn what was working and where. “They’re like, ‘Oh my God, I’m big in the U.K.! I never knew I had a London following!’” says Walk. Superusers such as Cruikshank were so successful in exploiting YouTube’s business initiatives that corporations such as Sony were studying their methodology and even paid some of them consultant fees to help them understand the digital world.

  The dynamic between Kamangar and Hurley was interesting to watch. Hurley was still YouTube’s CEO, while Kamangar considered himself “a facilitator.” (YouTube cofounder Steve Chen left his role as YouTube’s chief technology officer in June 2009, still working at Google on various engineering projects.) Kamangar loved to stay behind the scenes; he had to be dragged to a magazine photo shoot that paired him with Hurley in the lead photo. But Hurley clearly respected Kamangar’s opinions to the point of deference. Consider the behavior in a meeting of YouTube’s lead managers one day in 2009. Under discussion was th
e question of when YouTube should show its videos in high definition. Brin had sent word that he was pushing for it. But it would be expensive for YouTube to stream those bigger files over the network. Kamangar held back during the discussion of costs until the room paused to allow him to speak. “I thought we were going to stay within our budget,” he said. That deflated the advocates for making HD the default mode for playback. YouTube’s engineering director suggested that maybe they could just test making HD the automatic choice, so YouTube could measure the impact. Kamangar wondered whether such a choice would shift users’ expectations, making it impossible to reverse the move. Finally he suggested a compromise: Google would stream HD as the first option in the professionally produced, copyrighted videos known as “partner content.” Later, as Google delivered more broadband, it would focus on countries where a larger percentage of videos produced ad revenue. As he made this suggestion, Hurley and the rest nodded in agreement, and that was it. (Kamangar would become the official CEO of YouTube in October 2010, with Hurley assuming an advisory role.)

  YouTube didn’t become profitable in 2009, but it was making back enough of its expenses for Google executives to consider Kamangar’s tenure successful. Advertisers were paying for a billion “monetized views” a week. “We’ve done an incredible job in bringing costs down and revenues up,” Kamangar says. “It’s obvious that the basic model is correct.”

  In September 2009, the top executives of Google decisively agreed that was the case. YouTube’s leaders ventured down to Mountain View for a GPS meeting, and after assessing the numbers, the verdict was unanimous: YouTube had made it. “Basically things are fine,” said Eric Schmidt. “You’re at the point where the question is what to do next.”